@sallmarta/eye-hate-agent 1.0.2 → 1.0.4

package/docs/project-docs/changelog.md

@@ -1,299 +0,0 @@
-# Changelog — Eye Hate Agent
-
-All notable changes to Eye Hate Agent are documented here. Keep in mind, `docs/project-docs/changelog.md` has to be updated whenever important things change in this repository.
-
-## [1.18.2] - 2026-05-26
-
-### Changed
-
-- **Bare scoped npx follow-up**: Reordered the package bin entries and bumped `package.json` to `1.0.2` so bare `npx @sallmarta/eye-hate-agent ...` should prefer `eye-hate-agent` before the installed `eha` alias.
-
-## [1.18.1] - 2026-05-26
-
-### Changed
-
-- **npx entrypoint compatibility**: Added a package-name bin alias (`eye-hate-agent`) alongside `eha` and bumped `package.json` to `1.0.1` so `npx @sallmarta/eye-hate-agent ...` resolves correctly after republishing.
-
-## [1.18.0] - 2026-05-26
-
-### Added
-
-- **MIT license file**: Added a root `LICENSE` file so the published npm package has an explicit license artifact.
-
-### Changed
-
-- **Scoped package identity**: Updated `package.json` to use the public scoped package name `@sallmarta/eye-hate-agent`.
-- **npm metadata hardening**: Added repository, homepage, bugs URL, keywords, author, `publishConfig.access`, and a Node.js engine requirement for npm publishing.
-- **README install guidance**: Updated `README.md` to distinguish local development, global install, and one-off `npx` usage for the scoped package.
-- **Packaged CLI portability**: Updated the engine to accept generic project roots and materialize bundled contract and workflow sources into `.eha/generated/sources/` so installed tarballs work outside the EHA source repo.
-- **Release workflow docs**: Updated `docs/project-docs/foundation/workflow.md` and `docs/project-docs/testing.md` with package validation and publish-preparation steps.
-
-## [1.17.0] - 2026-05-25
-
-### Added
-
-- **EHA engine foundation**: Added `src/engine/` with a workflow registry, repo-root/state handling, runtime adapters, install/uninstall orchestration, and workflow prompt preparation.
-- **Engine-backed CLI**: Replaced the thin prompt-printing CLI with an engine-backed `eha` control plane supporting `install`, `uninstall`, `doctor`, `bootstrap`, `refresh`, `parity`, `discuss`, `execute`, `verify`, and `init`.
-- **Repo-local runtime projections**: Added generated runtime outputs for Claude (`.claude/commands/eha/`) and GitHub Copilot (`.github/instructions/eha-workflows.instructions.md`).
-- **Owner docs for the engine**: Added `docs/project-docs/index.md`, `foundation/architecture.md`, `foundation/workflow.md`, and `foundation/status.md` to own the new engine direction.
-- **Engine tests**: Added a Node built-in test suite covering workflow dispatch, repo-root discovery, and install/uninstall behavior.
-- **Newest workflow reference**: Added `newest-flow.md` at the repository root with full current-state flowcharts and step-by-step workflows for new projects and running/mature projects.
-
-### Changed
-
-- **Repository identity**: Updated `README.md` so EHA is described as a template-and-engine repository instead of template-only.
-- **Verification policy**: Updated `docs/project-docs/testing.md` so engine and CLI changes require executable checks in addition to the existing document-first maintenance review.
-- **Maintenance scope**: Updated `docs/eyehateagent-maintenance.md` so engine code, installer behavior, and runtime projections are explicitly part of template-repo maintenance.
-
-## [1.16.0] - 2026-05-24
-
-### Changed
-
-- **Parity Fix — `guidelines/` → `technical-guidelines/`**: Standardized all references across the contract, README, maintenance doc, template `technical-guidelines/index.md`, `full-verification` skill, and `test-authoring` skill to use `technical-guidelines/` as the canonical subdirectory name. Eliminates the most pervasive naming inconsistency in the repository.
-- **Parity Fix — flat paths → 4-layer paths**: Updated all stale flat-path references (`project.md`, `architecture.md`, `testing.md`, `status.md`) to use 4-layer taxonomy paths (`foundation/prd.md`, `foundation/architecture.md`, `technical/testing.md`, `foundation/status.md`) across the contract (Naming And Surface Rules, Request Routing Examples, Registry Policy, Guideline Policy, Adoption Checklists), README (required docs list, recommended docs), and all seven skills.
-- **Parity Fix — drop `quick-reference.md`**: Removed all ghost references to `quick-reference.md` from the contract (Naming And Surface Rules, Adoption Checklists), and from the Required Project Inputs tables of all seven skills (`parity`, `full-verification`, `analysis`, `api-design`, `code-audit`, `project-elevation`, `test-authoring`).
-- **Parity Fix — `project.md` → `prd.md`**: Updated all remaining references to the absorbed `project.md` to `prd.md` or `foundation/prd.md` across skills and the contract.
-- **Parity Fix — maintenance doc headings**: Added numbered prefixes to all section headings in `docs/eyehateagent-maintenance.md` to match the Stable Headings defined in the contract.
-- **Parity Fix — file casing**: Renamed `docs/project-docs/CHANGELOG.md` → `changelog.md` and `docs/project-docs/TESTING.md` → `testing.md` to match the contract's lowercase convention.
-- **Parity Fix — CLI and package**: Fixed `bin/eha.js` version from `2.0.0` to `1.0.0` to match `package.json`. Added missing `commander` and `chalk` dependencies to `package.json`. Removed stale `"main": "index.js"` entry.
-- **Parity Fix — template guideline paths**: Updated `docs/vibes/project-docs-template/technical-guidelines/index.md` guideline file paths from `guidelines/` to `technical-guidelines/`.
-- **Parity Fix — testing.md starter path**: Updated `docs/project-docs/testing.md` to reference the correct 4-layer starter template path.
-
----
-
-## [1.15.0] - 2026-05-24
-
-### Added
-
-- **EHA CLI Tool**: Introduced `bin/eha.js` and a `package.json` to make Eye Hate Agent an executable Node module. The CLI allows users to initialize and run parity/SDD prompts directly via terminal commands.
-- **Spec-Driven Development (SDD) Prompts**: Created brand new executable AI workflows `01-sdd-execute.md` (for TDD/SDD code generation) and `02-sdd-discuss.md` (for brainstorming specifications before writing code) to formalize the agile execution loop.
-
-### Changed
-
-- **Massive Template Overhaul**: Restructured `docs/vibes/project-docs-template` into a strict 4-layer taxonomy (`foundation`, `operations`, `technical`, `technical-guidelines`).
-- **Mega PRD Integration**: Absorbed `project.md` and `business-process.md` into `foundation/prd.md` to prevent context fragmentation.
-- **Testing Consolidation**: Absorbed `test-plan.md` and `test-case.md` into `technical/testing.md` to streamline the verification policy.
-- **Data Dictionary**: Formalized a Data Dictionary section inside `technical/database.md`.
-- **Backlog Unified**: Designated `foundation/phases.md` as the unified agile sprint tracker, officially replacing `sprints.md`.
-- **Contract Sync**: Completely rewrote `eyehateagent-contract.md` and `README.md` to align the "Required Document Set", "Stable Headings", and "Ownership Rules" with the new 4-layer architecture.
-- **Reusable Prompts Sync**: Rewrote the Output Contracts and target paths across all executable AI workflows (`00-project-docs-bootstrap.md`, `01-sdd-execute.md`, `02-sdd-discuss.md`, `00-project-docs-parity.md`, `00-project-docs-refresh.md`) to strictly enforce the new 4-layer taxonomy and eliminate hardcoded legacy folders.
-
----
-
-## [1.14.0] - 2026-05-22
-
-### Changed
-
-- Enhanced Rule 2.5 (Claude) and Rule 2.4 (Gemini, Copilot) for Session Continuity across all platform instruction surfaces. Standardized the handoff filename to a predictable, overwriting `session-handoff.md` under `active-repo/memories/session/session-handoff.md` to prevent directory clutter. Explicitly defined handoff compilation requirements to mandate a comprehensive active conversation summary, strict redaction of sensitive data (API keys, credentials, PII), dynamic tailoring based on user-provided compaction arguments, and providing a copy-pasteable short prompt (e.g., "Resume session from memories/session/session-handoff.md") for seamless onboarding.
-
----
-
-## [1.13.0] - 2026-05-22
-
-### Changed
-
-- Consolidated all three platform instruction surfaces (`.claude/rules/agent-rules.md`, `.agents/rules/agent.md`, and `.github/instructions/agent-rules.instructions.md`) to a point-numbered (`#1.1`, `#2.4`, etc.) 5-heading schema to achieve perfect consistency and remove redundancy.
-- Replaced the generic `[AGENT-SPECIFIC CACHE RULE GOES HERE]` placeholder with researched, platform-specific cache-integrity rules:
-  - **Gemini** (`.agents/rules/agent.md`): Added `Prefix Stability` rule targeting implicit prefix cache protection, byte-identical prefix requirements, and non-deterministic serialization risks.
-  - **Claude** (`.claude/rules/agent-rules.md`): Added `Prefix & Lookback Integrity` rule targeting byte-identical prefix order, dynamic injection prevention, and the ~20-block lookback window overflow.
-  - **Copilot** (`.github/instructions/agent-rules.instructions.md`): Added `Context Efficiency` rule targeting usage-based AI Credits billing, explicit over implicit context, and session hygiene.
-- Merged redundant rule **4.1** (`Keep rules generic...`) into **2.1** (`Context Limit`) under `## 2. Context & Cache Integrity`, cleanly shift-renumbering all Section 4 items to keep the completion section tightly focused.
-- Strengthened **Section 5 (Contract Essentials)** across all surfaces with bolded subtitles and advanced behavioral safeguards:
-  - Formally embedded the nested **Decision Precedence** list directly from the contract.
-  - Upgraded **Failure & Fallback** to explicitly address code-doc drift (the "do not guess" rule on code, tests, and configs).
-  - Introduced **Codebase & Comment Integrity** to strictly prevent the silent deletion of comments, docstrings, formatting, and structures.
-  - Elevated **Completion & Verification** to actively search for correctness, edge cases, and unexpected side effects.
-
----
-
-## [1.12.0] - 2026-05-18
-
-### Added
-
-- Added `## Guidelines` as a stable heading to the `architecture.md` template and contract to ensure domain-specific rules are properly delegated to `guidelines/index.md`.
-- Added a note to `README.md` clarifying that adopters only need to keep the platform instruction surface(s) they actually use (typically `.agents` for Gemini, `.claude` for Claude, or `.github` for GitHub Copilot).
-- Added `.agents/rules/` to the adoption topology folder trees in `README.md` to match the newly supported 3-mirror architecture.
-- Added explicit, punchy "Pros and Cons" summaries to all three topology scenarios in `README.md` to help adopters choose the right architecture (including strict target isolation benefits for Scenario 3).
-
-### Changed
-
-- Unified repository taxonomy by changing all scattered references of "mirrored rule files" and "rules" to "platform instruction surfaces (mirrored rule files)" across the contract, maintenance doc, README, reusable prompts, and skills.
-- Restored the explicit `docs/eyehateagent-contract.md` dependency to the `full-verification` skill so it can accurately route template-level consistency checks to `parity`.
-- Updated the testing diagnostic commands in `docs/project-docs/testing.md` to explicitly search the new `.agents` directory for consistency checking.
-- Streamlined `api-design`, `analysis`, `test-authoring`, `project-elevation`, and `code-audit` skills by removing redundant dependencies on the meta-contract, focusing them strictly on `docs/project-docs/` for normal work.
-
----
-
-## [1.11.0] - 2026-05-18
-
-### Added
-
-- Added `## 7. Contract Essentials` section to all three mirrored agent rule files (`.claude/rules/agent-rules.md`, `.github/instructions/agent-rules.instructions.md`, `.agents/rules/agent.md`). Extracts four critical behavioral rules from the contract that agents need on every task: skill invocation override, decision precedence order, core fallback behaviors, and completion requirements. This ensures agents follow these rules even when they don't open the full contract.
-
-### Changed
-
-- Refined all three agent rule files: tightened wording for knowledge preservation, format precedence, approach description, and merged scattered verification rules into a single completion bullet in § 7.
-
-### Removed
-
-- Removed redundant contract file-reading trigger (line 43) from all rule files — § 7 Contract Essentials now covers what agents would find there.
-- Removed template-maintainer guidance ("Keep reusable rules free of stack-specific commands...") from always-on rules — this belongs in the contract's authoring section, not in agent behavior rules.
-
----
-
-## [1.10.0] - 2026-05-16
-
-### Added in 1.10.0
-
-- Added explicit rules and guidance across the contract, reusable prompts, and agent rules for preserving valuable project-specific knowledge from legacy documentation (e.g., 'Decision Rationale') even if it lacks a standard template heading. Agents are now instructed to evaluate whether such information should become a new custom section in an existing document or a new separate file, and to ask the user if the best approach is ambiguous.
-- Added explicit rules for codebase-only knowledge discovery during documentation bootstrap and refresh. When no legacy docs exist, agents now inspect code, comments, configs, tests, and repository structure for valuable domain knowledge and surface it as custom sections or files, marking codebase-inferred facts with lower confidence until the user confirms them.
-
----
-
-## [1.9.0] - 2026-05-15
-
-### Changed in 1.9.0
-
-- Renamed the `consistency-audit` workflow family to `parity` across the reusable prompt, matching skill, routing docs, README guidance, and verification guidance so the user-facing and internal names stay aligned.
-- Retired the old `00-project-docs-consistency-audit.md` and `docs/vibes/skills/consistency-audit/` paths so only the canonical parity surfaces remain in the active template.
-
----
-
-## [1.8.0] - 2026-05-15
-
-### Changed in 1.8.0
-
-- Expanded the consistency-audit prompt and related contract and skill surfaces so audits can inspect current implementation evidence when code and docs disagree, and require an ask-the-user pause when authority between docs and code is not explicit.
-- Tightened `00-project-docs-refresh.md` so legacy or reference folders such as `docs-old/` and `docs-legacy/` can promote still-valid optional docs, active guideline sets, and phased-planning docs into `docs/project-docs/` instead of being left stranded in reference-only paths.
-- Clarified README Step 3 with explicit migration guidance and example prompts for promoting legacy `guidelines/` and `phases/` material during refresh-oriented adoption.
-- Added a narrow contract-level migration rule so reusable prompts promote justified legacy optional docs into the active owner-doc set and update the relevant registries.
-- Refined the contract, refresh prompt, consistency-audit prompt, and README so legacy docs are mapped by governed concern and content rather than by legacy folder or filename.
-- Added representative semantic-normalization examples such as `epic` or `roadmap` mapping to phased-planning owners and `protocol` or `standard` mapping to guideline owners when their content matches those concerns.
-- Added an explicit ask-the-user fallback pattern for ambiguous legacy mappings so agents do not silently guess when a legacy label may need to be preserved or may map to multiple active owners.
-
----
-
-## [1.7.0] - 2026-05-13
-
-### Added in 1.7.0
-
-- Added a README `Legend` section that defines repo-specific terms in plain language for adopters and maintainers.
-
-### Changed in 1.7.0
-
-- Made legacy-doc migration explicit across the README, contract, and reusable prompts by treating clearly named folders such as `docs-legacy/` as reference input instead of active owner docs.
-- Condensed README Step 3 into a smaller scenario guide and migration checklist that is easier for both humans and agents to scan.
-- Simplified more README wording outside the Legend so the main guide uses less contract-heavy language.
-- Merged the README navigation and registry-extension guidance into one shorter `Key Paths` section and removed redundant file-summary sections.
-- Rephrased the README explanation of regular docs versus guidelines with simpler language and clearer examples.
-- Reworked README Step 3 into a shorter tutorial-style flow with one explicit procedure for new repos, existing docs, and legacy-doc migration.
-
-### Removed in 1.7.0
-
-- Removed one more redundant recap from README Step 3 so the prompt-selection flow reads more directly.
-- Removed two advanced Legend terms and shortened the README guideline example list so the guide stays lighter for first-time readers.
-
----
-
-## [1.6.0] - 2026-05-12
-
-### Added in 1.6.0
-
-- Added explicit maintainer-facing anchors for where to register new optional regular doc types and guideline types.
-- Added a registry-driven extension model for optional regular docs and guideline types.
-- Added a starter `docs/vibes/project-docs-template/index.md` registry and retuned reusable prompts to treat registries as inventory sources of truth.
-- Added a local `docs/project-docs/testing.md` owner doc for template-maintenance validation rules.
-- Added a recommended starter guideline pack and a required `guidelines/index.md` when guideline files exist.
-- Added starter guideline templates and updated reusable prompts and maintainer docs to keep guideline guidance consistent.
-- Added a central skill-selection matrix in `docs/eyehateagent-contract.md` to separate analysis, audit, design, verification, elevation, and consistency work more clearly.
-- Added a new `full-verification` skill as the broad verification entry point for code, docs, contracts, architecture, quality, and project-health checks.
-
-### Changed in 1.6.0
-
-- Refreshed `README.md` with a table of contents, a clearer file index, and more explicit registry-driven extension entry points.
-- Clarified the contract split between core project docs and technical guideline docs.
-- Tightened `project-elevation` so it is explicitly forward-looking rather than a generic analysis wrapper.
-- Clarified `test-authoring` as a verification-strategy-first skill that chooses the right checks before writing tests.
-- Updated the contract and overlapping specialist skills so broad verification requests route cleanly to `full-verification` before choosing a single best specialist path.
-
----
-
-## [1.5.0] - 2026-05-12
-
-### Added in 1.5.0
-
-- Added a repository index to `README.md` and refreshed the adoption examples to show the new contract layout.
-
-### Changed in 1.5.0
-
-- Moved the canonical template anchors under `docs/` as `docs/eyehateagent-contract.md` and `docs/eyehateagent-maintenance.md`.
-- Retargeted the platform instruction surfaces (mirrored rule files), reusable prompts, skills, and changelog references to the new docs-anchor paths.
-- Refined the contract and maintenance wording to match the new docs-anchor model while preserving the 65% context-compaction exception.
-
----
-
-## [1.4.0] - 2026-05-11
-
-### Added in 1.4.0
-
-- Added template governance through `docs/eyehateagent-maintenance.md`.
-
-### Changed in 1.4.0
-
-- Moved the contract anchors to the repository root and retargeted live references.
-- Formalized Scenario 2 support in `docs/eyehateagent-contract.md` while keeping Scenario 3 outside the contract.
-- Turned `README.md` into the main operator guide with a chained adoption flow: topology, target project category, and starter reusable prompt.
-
-### Removed in 1.4.0
-
-- Removed standalone adoption-guide surfaces that are now absorbed into the root `README.md`.
-
----
-
-## [1.3.0] - 2026-05-10
-
-### Changed in 1.3.0
-
-- Simplified the platform instruction surfaces (mirrored rule files) and clarified response-shape precedence through `docs/eyehateagent-contract.md`.
-- Shifted template verification back to a document-first, human-reviewed maintenance flow.
-
-### Removed in 1.3.0
-
-- Removed outdated sample-oriented docs from the active template tree.
-
----
-
-## [1.2.0] - 2026-05-09
-
-### Changed in 1.2.0
-
-- Refined contract, testing, quick-reference, and maintenance docs for clearer ownership and consistency.
-- Updated `README.md` and platform instruction surfaces (mirrored rule files) to better match the template workflow and terminology.
-
----
-
-## [1.1.0] - 2026-05-08
-
-### Added in 1.1.0
-
-- Added the root `README.md` as the first human-facing adoption guide.
-- Added optional starter docs for `prd.md` and `production-runbook.md`.
-
-### Changed in 1.1.0
-
-- Clarified keep, replace, and remove rules for copying the template into target repositories.
-- Tightened starter-pack and reusable-prompt guidance for downstream doc management.
-
----
-
-## [1.0.0] - 2026-05-07
-
-### Added in 1.0.0
-
-- Initial release of the template foundation with project-doc owner files, mirrored agent rules, reusable prompts, and reusable skills.
-- Added the first downstream adoption guide and the starter scaffold under `docs/vibes/project-docs-template/`.
-- Added starter `getting-started.md` and `changelog.md` for the scaffold.
-
-### Changed in 1.0.0
-
-- Elevated reusable prompts as a first-class template surface and standardized skill documentation for clearer guidance.

package/docs/project-docs/foundation/architecture.md

@@ -1,117 +0,0 @@
-# EHA Architecture
-
-Last updated: 2026-05-26
-
-## Purpose
-
-Eye Hate Agent is now a **hybrid template-and-engine repository**.
-
-It still owns:
-
-- the contract
-- mirrored platform instruction surfaces
-- reusable prompts
-- reusable skills
-
-It now also owns:
-
-- a repo-local engine
-- a repo-local installer
-- runtime projection logic
-- an engine-backed CLI
-
-## Core architecture
-
-### Canonical source surfaces
-
-Human-authored source-of-truth assets stay in:
-
-- `docs/eyehateagent-contract.md`
-- `docs/eyehateagent-maintenance.md`
-- `docs/vibes/reusable-prompts/`
-- `docs/vibes/skills/`
-- `.claude/rules/`
-- `.github/instructions/`
-- `.agents/rules/`
-
-These are edited directly by maintainers.
-
-### Engine layer
-
-Engine code lives under `src/engine/` and currently owns:
-
-- workflow registry
-- repo-root and state-path resolution
-- runtime adapter registry
-- install and uninstall orchestration
-- workflow prompt preparation for CLI dispatch
-
-### CLI layer
-
-`bin/eha.js` is the control plane.
-
-It owns:
-
-- runtime installation
-- runtime uninstallation
-- doctor output
-- workflow dispatch for `bootstrap`, `refresh`, `parity`, `discuss`, `execute`, and `verify`
-
-## Runtime support tiers
-
-### Claude
-
-**Support tier:** full
-
-Generated outputs:
-
-- `.claude/commands/eha/eha-bootstrap.md`
-- `.claude/commands/eha/eha-refresh.md`
-- `.claude/commands/eha/eha-parity.md`
-- `.claude/commands/eha/eha-discuss.md`
-- `.claude/commands/eha/eha-execute.md`
-- `.claude/commands/eha/eha-verify.md`
-
-Claude receives generated command surfaces from the workflow registry.
-
-### GitHub Copilot
-
-**Support tier:** guided
-
-Generated outputs:
-
-- `.github/instructions/eha-workflows.instructions.md`
-
-Copilot currently receives generated instruction routing rather than slash-command surfaces. This is an intentional capability gap and should remain explicit until a stronger Copilot command surface is defined.
-
-## Generated outputs model
-
-Generated outputs stay **inside the repository** for transparency.
-
-Current generated paths:
-
-- runtime surfaces in `.claude/` and `.github/`
-- engine manifest in `.eha/generated/install-manifest.json`
-- bundled contract and workflow source copies in `.eha/generated/sources/`
-- execution artifacts in `.eha/state/`
-
-### Ownership rule
-
-- canonical source assets are edited by humans
-- generated runtime outputs are materialized by `eha install`
-- generated outputs may be reviewed directly, but durable behavior changes should be made in the owning source assets or engine code first
-
-## Workflow registry model
-
-The engine workflow registry maps CLI commands to canonical reusable prompts:
-
-| CLI command | Canonical source |
-| --- | --- |
-| `bootstrap` / `init` | `docs/vibes/reusable-prompts/00-project-docs-bootstrap.md` |
-| `refresh` | `docs/vibes/reusable-prompts/00-project-docs-refresh.md` |
-| `parity` | `docs/vibes/reusable-prompts/00-project-docs-parity.md` |
-| `discuss` | `docs/vibes/reusable-prompts/02-sdd-discuss.md` |
-| `execute` | `docs/vibes/reusable-prompts/01-sdd-execute.md` |
-| `verify` | `docs/vibes/reusable-prompts/00-project-docs-parity.md` |
-
-`verify` currently routes to parity until a dedicated verify workflow exists.

package/docs/project-docs/foundation/status.md

@@ -1,32 +0,0 @@
-# EHA Status
-
-Last updated: 2026-05-25
-
-## Current status
-
-EHA is in the early productization phase.
-
-### Completed in the current phase
-
-- established the EHA engine direction
-- chose initial runtimes: Claude and GitHub Copilot
-- chose generated outputs to remain inside the repository
-- added an initial engine layer under `src/engine/`
-- replaced the thin CLI with an engine-backed CLI
-- added repo-local install, uninstall, and doctor commands
-- added runtime projection scaffolding for Claude and Copilot
-
-### Not complete yet
-
-- runtime support remains uneven: Claude is fuller than Copilot
-- `verify` still routes to parity because there is no dedicated verify prompt
-- brownfield intelligence and deeper workflow execution are not implemented yet
-- release discipline is still early
-
-## Productization roadmap
-
-1. deepen the runtime adapter contract and support tiers
-2. add richer engine state and artifact handling
-3. implement stronger workflow execution behavior beyond prompt preparation
-4. add brownfield analysis and project-state detection
-5. harden validation and release flow

package/docs/project-docs/foundation/workflow.md

@@ -1,63 +0,0 @@
-# EHA Workflow
-
-Last updated: 2026-05-25
-
-## Purpose
-
-This file owns the maintainer workflow for developing the EHA template and the EHA engine together.
-
-## Maintainer loop
-
-1. Edit the owning source assets:
-   - contract and maintenance docs
-   - reusable prompts and skills
-   - mirrored platform instruction surfaces
-   - `src/engine/` and `bin/eha.js`
-2. Regenerate repo-local runtime surfaces with `eha install claude copilot`
-3. Inspect engine state with `eha doctor`
-4. Run verification for changed code and docs
-5. Update `docs/project-docs/changelog.md`
-
-## Workflow commands
-
-The current engine-backed CLI commands are:
-
-- `eha install`
-- `eha uninstall`
-- `eha doctor`
-- `eha bootstrap`
-- `eha refresh`
-- `eha parity`
-- `eha discuss`
-- `eha execute`
-- `eha verify`
-- `eha init`
-
-## Install model
-
-The installer is currently **repo-local**.
-
-It generates runtime outputs into the repository itself instead of external config roots. This is intentional for transparency during the current productization phase.
-
-## Release and publish loop
-
-The target public package identity is `@sallmarta/eye-hate-agent`.
-
-Use this order for release preparation:
-
-1. Update package metadata and changelog.
-2. Run `npm test`.
-3. Run `npm pack --dry-run` and inspect the package contents.
-4. Run `npm pack`, install the tarball into a clean temporary directory, and smoke-test the CLI.
-5. Run `npm login` against the `@sallmarta` npm account when a real publish is intended.
-6. Publish with `npm publish --access public`.
-
-## Editing rule
-
-When behavior changes:
-
-- update engine code or canonical prompt assets first
-- regenerate runtime outputs second
-- update owner docs third
-
-Do not treat generated runtime outputs as the primary design surface.

package/docs/project-docs/index.md

@@ -1,20 +0,0 @@
-# EHA Project Docs Index
-
-This repository now has two active roles:
-
-1. the reusable Eye Hate Agent template system
-2. the EHA engine and installer code that productizes the template
-
-## Active owner docs
-
-| Path | Purpose | Status |
-| --- | --- | --- |
-| `docs/project-docs/foundation/architecture.md` | Owns the EHA engine architecture, runtime support tiers, and generated-output model | Active |
-| `docs/project-docs/foundation/workflow.md` | Owns maintainer and development workflow for the engine and repo-local installs | Active |
-| `docs/project-docs/foundation/status.md` | Owns current implementation status and phased productization roadmap | Active |
-| `docs/project-docs/testing.md` | Owns validation policy for template and engine changes | Active |
-| `docs/project-docs/changelog.md` | Owns maintainer-facing change history | Active |
-
-## Optional docs
-
-No additional optional owner docs are active yet for the EHA engine beyond the files listed above.

package/docs/project-docs/testing.md

@@ -1,73 +0,0 @@
-# Testing & Verification
-
-Last updated: 2026-05-12
-
----
-
-## Verification Policy
-
-This repository is now a hybrid template-and-engine repository.
-This owner doc applies to the `eye-hate-agent` repository itself; downstream repositories should define their own local `docs/project-docs/testing.md`, typically starting from `docs/vibes/project-docs-template/testing.md`.
-Validation for template-maintenance work remains document-first, but engine and CLI changes now require executable checks where available.
-Use the narrowest check that can prove the changed contract, prompt, skill, template, installer, or CLI behavior is internally consistent.
-
-## Verification Matrix
-
-| Change type | Preferred validation | Fallback |
-| --- | --- | --- |
-| Summary-only doc clarification | Markdown diagnostics on changed files, then targeted diff review | Manual read-through of the changed owner doc and dependent summary docs |
-| Contract, maintenance, reusable-prompt, or skill change | Markdown diagnostics, targeted consistency search across affected docs, then maintenance audit review | Structured manual consistency review across contract owner, dependent docs, and changelog |
-| Starter template change under `docs/vibes/project-docs-template/` | Markdown diagnostics on added or changed templates, then check that referenced template surfaces and indexes stay aligned | Manual folder inventory plus cross-reference review |
-| Canonical path, filename, stable heading, or routing change | Markdown diagnostics, consistency search across docs and platform instruction surfaces, then full maintenance audit review | Manual owner-by-owner review only if search tooling is unavailable |
-| Engine, installer, or CLI change | `npm test`, targeted CLI smoke checks, then owner-doc consistency review | Manual CLI smoke check plus structural code review if a full test does not exist |
-| npm package metadata or release-surface change | `npm test`, `npm pack --dry-run`, tarball smoke install, then owner-doc consistency review | Manual package file review plus local CLI smoke checks |
-
-## Commands
-
-- changed-file diagnostics: use editor or Problems diagnostics for every changed Markdown file
-- targeted consistency search: `rg -n "<changed-term>|<path>|<heading>" README.md docs .agents .github .claude`
-- changed-doc review: `git --no-pager diff -- README.md docs`
-- folder inventory for scaffold changes: `rg --files docs/vibes/project-docs-template`
-- engine test suite: `npm test`
-- CLI smoke check: `node bin/eha.js doctor`
-- package contents preview: `npm pack --dry-run`
-- tarball creation: `npm pack`
-- tarball smoke install: install the generated `.tgz` into a clean temporary directory, then run `npx eha doctor` or the installed `eha doctor`
-
-Replace `<changed-term>|<path>|<heading>` with the concrete contract term, file path, or heading touched by the change.
-
-## Test Layers
-
-- File diagnostics: confirm changed Markdown files have no syntax or lint issues.
-- Owner-doc consistency: confirm the owning doc expresses the durable truth first.
-- Dependent-doc consistency: confirm summaries, reusable prompts, skills, and maintenance docs still match the owner.
-- Template-surface consistency: confirm starter templates and indexes match the advertised contract.
-- Maintenance-audit review: for higher-impact changes, run the same cross-surface review expected by the maintenance workflow.
-
-## Naming and File Conventions
-
-- Owner testing doc path: `docs/project-docs/testing.md`
-- Adoption starter for downstream repos: `docs/vibes/project-docs-template/technical/testing.md`
-- Changelog path for maintainer-facing changes: `docs/project-docs/changelog.md`
-- Template starter docs live under `docs/vibes/project-docs-template/`
-- Reusable prompts live under `docs/vibes/reusable-prompts/`
-- Skills live under `docs/vibes/skills/`
-
-## CI / Release Gates
-
-- Required before merge: changed Markdown files are diagnostic-clean.
-- Required before merge: owning docs and dependent docs are checked for consistency.
-- Required before merge: maintainer-facing template changes are recorded in `docs/project-docs/changelog.md`.
-- Required before merge: engine and CLI changes pass `npm test`.
-- Required before publish: package metadata changes pass `npm pack --dry-run` and a tarball smoke-install check.
-- Required before release of breaking template changes: run the full maintenance audit sequence described by `docs/eyehateagent-maintenance.md`.
-- Advisory: confirm platform instruction surfaces (mirrored rule files) were either updated or explicitly left unchanged because they remain generic.
-
-## Manual Checks
-
-- Confirm the owning document was updated before dependent summaries or reusable assets.
-- Confirm contract changes did not leave stale paths, headings, or ownership rules in README, reusable prompts, maintenance docs, or skills.
-- Confirm new starter templates use the stable heading pattern defined by the contract.
-- Confirm the repo is not claiming executable tests that do not actually exist.
-- Confirm engine-generated runtime outputs still map back to canonical source assets and owner docs.
-- Confirm package metadata points at the canonical GitHub repository and intended npm scope.